sync ceno master with aot enable

.github/workflows/run-benchmark-v2.yml

@@ -95,10 +95,14 @@ on:
           - "1"
           - "2"
       max_cell_per_shard:
-        description: "Value for CENO_MAX_CELL_PER_SHARD; 1073741824 = (1 << 30) * 8 / 4 / 2"
+        # Magic number: the old shard cap was 1207959552 = ((1 << 30) * 9 / 4 / 2).
+        # Keccak preflight cost is now estimated with a 33/16 blowup instead of
+        # the old coarse 2x factor, so raise the shard cap by (33/16) / 2 = 33/32:
+        # 1207959552 * 33 / 32 = 1245708288.
+        description: "Value for CENO_MAX_CELL_PER_SHARD; magic default 1245708288 = 1207959552 * 33 / 32"
         required: false
         type: string
-        default: "1073741824"
+        default: "1245708288"
       gpu_jagged_reshape_log_height:
         description: "Value for CENO_GPU_JAGGED_RESHAPE_LOG_HEIGHT; 25 probably OOMs on RTX 4090"
         required: false
@@ -234,7 +238,7 @@ jobs:
           CENO_GPU_JAGGED_RESHAPE_LOG_HEIGHT=${{ inputs.gpu_jagged_reshape_log_height }} \
           CENO_GPU_CACHE_LEVEL=${{ inputs.gpu_cache_level }} RUSTFLAGS="-C target-feature=+avx2" \
           RUST_LOG=info,openvm_stark_*=warn,openvm_cuda_common=warn \
-          cargo run --features "jemalloc,gpu" --config net.git-fetch-with-cli=true \
+          cargo run --features "jemalloc,gpu,aot" --config net.git-fetch-with-cli=true \
             --release --bin ceno-reth-benchmark-bin -- \
             --mode prove-stark \
             --block-number $block_number \
@@ -396,7 +400,7 @@ jobs:
           CENO_GPU_JAGGED_RESHAPE_LOG_HEIGHT=${{ inputs.gpu_jagged_reshape_log_height }} \
           CENO_GPU_CACHE_LEVEL=${{ inputs.gpu_cache_level }} RUSTFLAGS="-C target-feature=+avx2" \
           RUST_LOG=info,openvm_stark_*=warn,openvm_cuda_common=warn \
-          cargo run --features "jemalloc,gpu" --config net.git-fetch-with-cli=true \
+          cargo run --features "jemalloc,gpu,aot" --config net.git-fetch-with-cli=true \
             --release --bin ceno-reth-benchmark-bin -- \
             --mode prove-stark \
             --block-number $block_number \

bin/ceno-host/Cargo.toml

@@ -22,7 +22,7 @@ metrics = ["openvm-reth-benchmark/metrics"]
 evm-verify = ["openvm-reth-benchmark/evm-verify"]
 perf-metrics = ["openvm-reth-benchmark/perf-metrics"]
 tco = ["openvm-reth-benchmark/tco"]
-# aot = ["openvm-reth-benchmark/aot"]
+aot = ["openvm-reth-benchmark/aot"]
 mimalloc = ["openvm-reth-benchmark/mimalloc"]
 jemalloc = ["openvm-reth-benchmark/jemalloc"]
 nightly-features = ["openvm-reth-benchmark/nightly-features"]

crates/host-bench/Cargo.toml

@@ -58,7 +58,7 @@ openvm-backend = []
 gpu = ["ceno_cli/gpu"]
 metrics = []
 tco = []
-# aot = ["openvm-sdk/aot"]
+aot = ["ceno_zkvm/aot-x86_64"]
 evm-prove = ["halo2-axiom"]
 evm-verify = ["evm-prove", "halo2-axiom"]
 perf-metrics = []

server/prove_block.sh

@@ -12,7 +12,11 @@ VERIFIER_ID="${VERIFIER_ID:-0.1}"
 CENO_GPU_CACHE_LEVEL="${CENO_GPU_CACHE_LEVEL:-1}"
 CENO_GPU_ENABLE_WITGEN="${CENO_GPU_ENABLE_WITGEN:-0}"
 CENO_CONCURRENT_CHIP_PROVING="${CENO_CONCURRENT_CHIP_PROVING:-1}"
-CENO_MAX_CELL_PER_SHARD="${CENO_MAX_CELL_PER_SHARD:-1073741824}"
+# Magic number: the old shard cap was 1207959552 = ((1 << 30) * 9 / 4 / 2).
+# Keccak preflight cost is now estimated with a 33/16 blowup instead of the old
+# coarse 2x factor, so raise the shard cap by (33/16) / 2 = 33/32:
+# 1207959552 * 33 / 32 = 1245708288.
+CENO_MAX_CELL_PER_SHARD="${CENO_MAX_CELL_PER_SHARD:-1245708288}"
 CENO_GPU_JAGGED_RESHAPE_LOG_HEIGHT="${CENO_GPU_JAGGED_RESHAPE_LOG_HEIGHT:-23}"
 CENO_GPU_LARGE_TASK_BOOKING_MARGIN_MB="${CENO_GPU_LARGE_TASK_BOOKING_MARGIN_MB:-3048}"
 RUST_MIN_STACK="${RUST_MIN_STACK:-536870912}"